Changed the paradigm for distribution_by, and introduced aggregate_by.

m.aggregate_by(:users) looks like
{user_id => number of tracks by this user}

m.distribution_by(:users) is now equivalent to
m.aggregate_by(:users).histogram

lib/modesty/datastore/redis.rb

@@ -77,22 +77,23 @@ def distribution(date)
       end
 
       def unique(param, date)
-        data.scard(key_for_with(param, date))
+        data.hlen(key_for_with(param, date))
       end
 
       def all(param, date)
-        data.smembers(key_for_with(param, date)).map(&:to_i?)
+        data.hkeys(key_for_with(param, date)).map(&:to_i?)
       end
 
       def distribution_by(param, date)
-        ids = data.smembers(key_for_with(param, date)).map(&:to_i?)
-        h = {}
-        ids.each do |id|
-          h[id] = Hash[data.hgetall(key_for_with(param, date, id)).map do |k,v|
-            [k.to_i?, v.to_i]
-          end]
-        end
-        return h
+        data.hvals(self.key_for_with(param, date)).map(&:to_i?).histogram
+      end
+
+      def aggregate_by(param, date)
+        Hash[
+          data.hgetall(self.key_for_with(param, date)).map do |k,v|
+            [k.to_i?, v.to_i?]
+          end
+        ]
       end
 
       def track!(count, with_args)
@@ -113,9 +114,7 @@ def add_counts(date, count)
       end
 
       def add_param_counts(date, count, param, id)
-        data.sadd(key_for_with(:__keys__), param)
-        data.sadd(key_for_with(param, date), id)
-        data.hincrby(key_for_with(param, date, id), count, 1)
+        data.hincrby(key_for_with(param, date), id, count)
       end
 
       def count_unidentified_user(date)

lib/modesty/metric/data.rb

@@ -27,7 +27,10 @@ def parse_date_or_range(start=nil,fin=nil)
     end
 
 
-    [:count, :distribution].each do |data_type|
+    [
+      :count,
+      :distribution
+    ].each do |data_type|
       data_type_by_range = :"#{data_type}_by_range"
       define_method(data_type) do |*dates|
         date_or_range = parse_date_or_range(*dates)
@@ -47,7 +50,12 @@ def parse_date_or_range(start=nil,fin=nil)
       end
     end
 
-    [:all, :unique, :distribution_by].each do |data_type|
+    [
+      :all,
+      :unique,
+      :distribution_by,
+      :aggregate_by
+    ].each do |data_type|
       by_range = :"#{data_type}_by_range"
       define_method(data_type) do |sym, *dates|
         sym = sym.to_sym

spec/identity_spec.rb

@@ -108,7 +108,9 @@
 
     @group_exp.distribution.should == {45 => 1, 30 => 1, 15 => 1}
     @group_ctrl.distribution.should == {}
-    @group_exp.distribution_by(:creators).should == {700 => {45 => 1, 30 => 1, 15 => 1}}
+    @group_exp.aggregate_by(:creators).should == {700 => 90}
+    @group_ctrl.aggregate_by(:creators).should == {}
+    @group_exp.distribution_by(:creators).should == {90 => 1}
     @group_ctrl.distribution_by(:creators).should == {}
 
     @group_exp.all(:creators).should == [700]

spec/metric_spec.rb

@@ -155,9 +155,12 @@
     m.unique(:zings, Date.parse('1/1/2002')).should == 0
     m.unique(:zings, :all).should == 2
 
-    m.distribution_by(:zings).should == {56=>{1=>1}, 97=>{7=>1, 4=>1}}
     m.distribution.should == {1 => 1, 7 => 1, 4 => 1}
 
+    #one zing has one track, and one zing has 11 (= 4+7) tracks.
+    m.distribution_by(:zings).should == {1 => 1, 11 => 1}
+    m.aggregate_by(:zings).should == {56 => 1, 97 => 11}
+
     m.distribution(:all).should == {1 => 1, 7 => 1, 4 => 1}
   end
 end